Datasets:

---

e-dealba

/

M2S2

Like 1

Hi Enrique and the M2S2 team,

Thank you for sharing M2S2. I have read your paper and downloaded the dataset, and I would appreciate some clarification on how to access the individual spectral bands in the HIL data.

Section 2.1 describes the TOUCAN camera as having a 2048 × 2048 raw resolution, a 512 × 512 spectral resolution, and 10-bit digitization. It also mentions using COLOR SHADES LAB SDK for spectral normalization, crosstalk correction, and hypercube extraction. Sections 2.4.1 and 2.4.2 describe using bands 5, 9, and 10 for the HIL segmentation experiments.

However, I have not been able to identify how the released HIL files correspond to these individual spectral bands.

For example, the files all_0.png, dir_0.png, and solo_0.png in HIL/Hubble/b1/e0/ are 2048 × 2048, 8-bit RGB PNGs. In the files I checked, the three stored channels are identical, and a repeating 4 × 4 pattern is visible.

Could you help clarify the following points?

1. Accessing the ten HIL spectral bands

Are these PNGs exports of the sensor mosaic before spectral-band extraction, or do they represent a different processing stage?

Are the extracted 512 × 512 spectral-band images or ten-band cubes available separately? If so, could you provide their location, band order, and central wavelengths?

In particular, how can I obtain the HIL bands 5, 9, and 10 used in the paper from the released files?

2. Decoding and calibration

If spectral-band extraction is intended to be performed from the published PNGs, could you share the mosaic/filter-position mapping and the required processing steps?

Would this require camera-specific calibration or crosstalk correction files? A minimal decoding example, or one input frame paired with its verified ten-band output, would be very helpful.

If the PNGs do not retain sufficient information for this processing, are the original higher-bit-depth captures available?

3. Processing applied before release

Section 2.3 describes dark-frame subtraction followed by clipping negative values to zero. Has this already been applied to the released HIL PNGs?

Could you also clarify how the camera data were converted to 8-bit PNGs, including any intensity scaling, gamma correction, clipping, or other transformations? Were spectral normalization and crosstalk correction applied before export?

What do the resulting pixel values represent: digital counts, corrected intensities, radiance, or reflectance?

4. Capture naming and metadata

The paper describes directional, weak diffuse, and strong diffuse illumination. Could you confirm how these correspond to the dir, solo, and all filename prefixes?

Section 2.3 also describes acquisition at 10-degree intervals about the z-axis and at positive, flat, and negative tilts. How do e0, e1, and e2 map to these tilt settings, and how do the frame indices map to rotation angles?

If camera intrinsics, distortion parameters, or camera/object pose metadata are available, could you point me to them as well?

My main question is how to go from the released HIL PNGs to the individual spectral bands used in the paper. Any processing code, documentation, or worked example would be greatly appreciated.

Thank you for your help and for making the dataset available.